Generally, digital conference arrangements combine input samples such as, for example, a speech sample from each conference channel and then distribute the resulting output sample to each conferee. With such time division multiplexed arrangements, the quality of speech communication between conferees should approximate that of a two-party connection. However, this is generally not the case, because the conferees usually communicate with each other over various types of transmission facilities which introduce noise, echo, and other types of distortion.
One of the more common ways to reduce the distortion is to include only the speech samples from selected speakers in the resulting output sample. Typically, this is accomplished by including in the conference circuit a detector circuit to designate a channel which has speech samples above a threshold level which is usually somewhat higher than the level of noise on the transmission facility. In response, conference circuit switching hardware includes only speech samples from designated channels in the resulting output sample. Consequently, noise and other distortion from non-speaking conferees is eliminated from or, in effect, switched out of the distributed sample. However, some minimum response time is required to detect speech samples above the threshold and to select designated channels. Depending on the amount of delay, the response time may cause the first syllable or even the entire first word of a new speaker to be excluded or, more commonly, clipped from the output signal. Since speech clipping increases directly with the number of conferees, the need to reduce or eliminate the clipping associated with large numbers of conferees becomes more pronounced so as to maintain intelligible communication.